Jeffrey Allen McDonald (“McDonald”) pled guilty to two counts of attempting to manufacture methamphetamine in violation of 21 U.S.C. §§ 846 and 841(b)(1)(B) and one count of creating a substantial risk of harm to human life while manufacturing a controlled substance in violation of 21 U.S.C. § 858. After calculating an advisory United States Sentencing Commission Sentencing Guidelines sentencing range of 262 to 327 months, the district court sentenced McDonald to 132 months’ imprisonment — 12 months longer than the statutory mandatory minimum sentence and roughly 50 percent below the guidelines sentencing range. The Government appeals. We remand for resentencing.
I. BACKGROUND
On January 2, 2004, law enforcement officers arrested McDonald after discovering that he was operating a methamphetamine lab in a crowded residential trailer park. On April 19, 2004, McDonald was arrested again. This time, during a traffic stop, police found him in possession of drug paraphernalia and materials necessary to manufacture methamphetamine. A subsequent search of McDonald’s residence yielded additional methamphetamine manufacturing materials. McDonald faced state charges in Iowa following each arrest.
The state charges against McDonald were dismissed when a federal grand jury indicted McDonald in connection with his January and April 2004 arrests. In the district court, McDonald admitted his January 2004 offense and pled guilty to one count of attempting to manufacture methamphetamine. Because his methamphetamine lab created a serious risk of explosion in the crowded trailer park, McDonald also pled guilty to one count of creating a substantial risk to human life while manufacturing a controlled substance. Related to his April 2004 arrest, McDonald pled guilty to a second count of attempting to manufacture methamphetamine.
Because McDonald had previously been convicted of a felony drug offense, the district court determined that the statutory mandatory minimum sentence for each count of attempting to manufacture methamphetamine was ten years. 21 U.S.C. § 841(b)(1)(B). The maximum available sentence was life imprisonment. Id. Because the district court determined that McDonald was a “career offender,”1 McDonald’s criminal history category was VI. Guidelines § 4B1.1. Because the statutory maximum sentence for his crime was life imprisonment, McDonald’s adjusted base offense level was 37. Id. McDonald received a three-level adjustment for acceptance of responsibility. Accordingly, the district court determined that the total offense level was 34. As a result, the advisory guidelines sentencing range was 262 to 327 months.2
*951At McDonald’s sentencing hearing, McDonald’s brother, Rodney McDonald (“Rodney”), testified that after McDonald was paroled from prison in 1998, McDonald stayed off drugs, worked as a mechanic and won custody of his two daughters. According to Rodney, in 2001, one of McDonald’s daughters was killed in an automobile accident, and McDonald’s other daughter moved back in with her mother. Rodney testified that McDonald began drinking again, and he suggested that McDonald resumed his drug use. Rodney also stated that following McDonald’s prison term for the instant offenses, McDonald’s family would be supportive, and McDonald’s most recent employer would be willing to rehire McDonald “if he’s able to stay off drugs.”
Following Rodney’s testimony, the district court entertained McDonald’s attorney’s argument that: (i) the death of McDonald’s daughter, though not “an excuse, ... is clearly relevant as to how Mr. McDonald got back into using meth;” (ii) McDonald manufactured methamphetamine “primarily to feed his habit;” (in) McDonald’s criminal history was overstated due to the career offender enhancement; (iv) McDonald is only 38 years old, has a “pretty good work history” and has strong family support; and (v) McDonald began cooperating “very early on” after his arrest. McDonald’s lawyer argued that McDonald should be sentenced to the statutory mandatory minimum sentence of 120 months.
McDonald testified that he had been “mistreated in the system,” he had already served time when his probation had been revoked in 1997, and “he stayed clean” beginning with his imprisonment in 1997 and up until his daughter died in 2001.
After also hearing from the Government, the district court discussed factors that, in its view, were relevant to McDonald’s sentence. The district court viewed positively McDonald’s “employment for over a year’s period of time as well as his skills as a welder,” finding that “when [McDonald] is not feeding his addiction, he can compete in the private sector.” The district court considered the statement of McDonald’s previous employer, Ms. Schulte of Schulte Construction. Schulte Construction employed McDonald “as a mechanic and a laborer from March 17, ’02 through June 28, ’03.” Ms. Schulte told the probation office that McDonald “was an excellent employee, he was very knowledgeable, hardworking and punctual.” However, Ms. Schulte also “relate[d] how. [McDonald’s] productivity and attendance declined.” The district court stated that it “assumefd] this is in conjunction with his addiction.” Ms. Schulte indicated that she might reconsider McDonald for employment following his imprisonment, if he can stay off drugs.
On the other hand, the district court found that McDonald was appropriately considered a career offender, specifically disagreed with McDonald’s “characterization of his past activity as being innocuous” and noted that McDonald’s activities placed the public at serious risk.
The district court made several comments about McDonald’s psychological health and drug addiction. For instance, the district court noted that McDonald had been diagnosed with depression and was taking medication. The district court also stated that “[s]pecific mention by me will be [made] of the Defendant’s long-time drug dependency and psychiatric diagnoses that have been made both in the past and in the future” and that “[t]he Defendant’s criminal history indicates that he has been addicted to methamphetamine.” The district court concluded that *952it was “pretty well convinced that the Defendant is a drug addict and has a great dependency on drugs ...Nevertheless, the district court stated that McDonald’s addiction “does not forgive or relieve the serious implications of his drug manufacturing in this instance.”
The district court also addressed the death of McDonald’s daughter, stating that “[t]he personal tragedies of the Defendant have, according to the Defendant, been the triggering event. Whether or not that’s true, and the Court is in no position to know that, the Defendant apparently has no insight into his mental health problems, nor apparently does his family.”
Finally, the district court discussed an unpublished opinion from the Northern District of Indiana, United States v. Nellum, No. 2:04-CR-30-PS, 2005 WL 300073 (N.D.Ind. Feb.3, 2005). Citing a United States Sentencing Commission study regarding recidivism,3 the district court in Nellum varied downward from the guidelines sentencing range based in part upon that court’s prediction that Nellum was unlikely to recidivate due to his advanced age. Relying upon Nellum’s reproduction of a portion of a chart from the USSC Recidivism Study, the district court in the instant case stated that
[t]he likelihood of recidivism ... recidivism rates decline consistently as age increases. Generally, the younger the offender, the more likely the offender recidivates.... [Quoting Nellum: ] Among all offenders under the age of 21 recidivism rate is 35.5, while offenders over the age of 50 have a recidivism rate of 9.5 percent.... According to the [criminal] history category III Defendants like Defendant Nellum the recidivism rates are as follows, and the Court is going to specifically refer to Mr. McDonald’s age as between 36 and 40, the recidivism rate is still 29.4 percent.
Based upon the district court’s analysis of McDonald’s age and its perceived relationship to the likelihood that McDonald would recidivate, the district court stated that “with the long sentence that is mandated by the Congress, hopefully the citizens will be protected from Mr. McDonald. He will be adequately deterred and the public won’t have further crimes committed by this Defendant.”
The district court then sentenced McDonald to 132 months’ imprisonment on each count of attempting to manufacture methamphetamine, as well as 120 months’ imprisonment for creating a substantial risk of harm to human life while manufacturing a controlled substance. McDonald’s sentences for all counts were to run concurrently. The district court concluded by commenting that “[t]his ..., I think, would be categorized by the Government as a very lenient sentence .... ”
II. DISCUSSION
Following the Supreme Court’s decision in United States v. Booker, 543 U.S. 220, 125 S.Ct. 738, 160 L.Ed.2d 621 (2005), we review sentences to determine “whether the district court abused its discretion by imposing an unreasonable sentence on the defendant.” United States v. Haack, 403 F.3d 997, 1003 (8th Cir.2005). The guidelines sentencing range remains the “critical starting point” of our analysis, United States v. Mashek, 406 F.3d 1012, 1016 n. 4 (8th Cir.2005), and “[b]ecause the Guidelines are fashioned taking the other § 3553(a) factors into account and are the product of years of careful study, the *953guidelines sentencing range, though advisory, is presumed reasonable,” United States v. Shafer, 438 F.3d 1225, 1227 (8th Cir.2006).
Nevertheless, because the sentencing guidelines are advisory, “the [district] court has a range of choice, and ... its decision will not be disturbed as long as it stays within that range and is not' influenced by any mistake of law.” Haack, 403 F.3d at 1004 (quoting Kern v. TXO Prod. Corp., 738 F.2d 968, 970 (8th Cir.1984)). Critically, the “range of choice is limited.” Id. “[T]he farther the district court varies from the presumptively reasonable guidelines range, the more compelling the justification based on the § 3553(a) factors must be.” United States v. McMannus, 436 F.3d 871, 874 (8th Cir.2006); see also United States v. Dalton, 404 F.3d 1029, 1033 (8th Cir.2005) (holding in review of a downward departure that “[a]n extraordinary reduction must be supported by extraordinary circumstances”). A district court abuses its sentencing discretion if it
fails to consider a relevant factor that should have received significant weight, gives significant weight to an improper or irrelevant factor, or considers only appropriate factors but nevertheless commits a clear error of judgment by arriving at a sentence that lies outside the limited range of choice dictated by the facts of the case.
Haack, 403 F.3d at 1004.
The record reflects that the district court clearly articulated two reasons in support of the 130-month variance it awarded McDonald: McDonald’s work history and the district court’s estimation of McDonald’s likelihood of recidivism. We find that neither reason is sufficient to support the extraordinary sentence reduction.
During sentencing, with respect to McDonald’s employment history, the district court only “reference[d] [McDonald’s] employment for over a year’s period of time [i.e., his 15 months with Schulte Construction] as well as his skills as a welder.” Even assuming that a 15-month period of work could be an extraordinarily positive factor when assessing a 38 year-old defendant’s work • history, it is clear that McDonald’s work history was anything but exemplary, even during that brief period. To the contrary, the record indicates that, while McDonald at first attended work regularly, his attendance severely declined due to his drug habit. As his employer, Ms. Schulte, recounted, McDonald “became very lax in coming to work on time, if he even showed up.” Moreover, there is no indication in the record that McDonald was-employed for the six months-leading up to his January 2004 arrest. Accordingly, to the extent that McDonald’s work during the 15-month period discussed by the district court contributed to the sentencing variance, it cannot sustain the extraordinary variance in this case. Cf. United States v. Moreland, 437 F.3d 424, 437 (4th Cir.2006) (“Moreland’s desultory pursuit of his education and his spotty employment history — six jobs over the course of seven years, with his last legitimate employment in 1999 — can provide little confidence in his .willingness to become a productive member of society, irrespective of his ability to do so.”).
The district court’s determination that McDonald was unlikely to recidivate following his abbreviated sentence — a determination that relied upon McDonald’s age and the USSC Recidivism Study- — is similarly insufficient to warrant a variance of the magnitude awarded. The guidelines have already accounted for the likelihood of recidivism through the guidelines’ criminal history computation. See, e.g., United States v. Gayle, 389 F.3d 406, 409 (2d Cir.2004) (“Gayle’s criminal history category ... serves as a proxy for his likelihood of recidivism.”); USSC Recidivism Study *954at 10 (“Both [Criminal History Categories] and criminal history points predict recidivism.”); id. at 13 (“[T]he criminal history computation is designed to predict recidivism.”). Indeed, the USSC Recidivism Study concluded that the guidelines’ criminal history accurately predicted the likelihood of recidivism. Id. at 15 (“Testing of the guidelines’ criminal history measure’s predictive powers shows that the aggregate Chapter Four provisions are performing as intended and designed.... The empirical evidence shows that criminal history as a risk measurement tool has statistically significant power in distinguishing between recidivists and non-recidivists.”).
We have previously cautioned against substantial variances predicated upon characteristics of the individual defendant for which the guidelines calculation already accounts. In United States v. Myers, 439 F.3d 415, 418 (8th Cir.2006), we rejected such a variance:
The only section 3553(a) factor identified by the district court that weighs toward a more lenient sentence is Myers’ lack of criminal history. Inasmuch as a guidelines sentence reflects a defendant’s criminal history, a wide divergence from the guidelines sentence based solely on this single criterion would conflict with the need to avoid unwarranted sentence disparities among defendants with similar records who have been found guilty of similar conduct.
See also United States v. Claiborne, 439 F.3d 479, 481 (8th Cir.2006) (rejecting a variance based upon the defendant’s criminal history and other factors for which the guidelines already accounted). As Myers noted, substantial variances based upon factors already taken into account in a defendant’s guidelines sentencing range seriously undermine sentencing uniformity. 439 F.3d at 418. And as Booker stated, Congress’s clear intent is to make sentences more uniform, and uniformity remains an important sentencing goal. 543 U.S. at 253-54, 125 S.Ct. 738; see also 18 U.S.C. § 3553(a)(6) (including “the need to avoid unwarranted sentence disparities among defendants with similar records who have been found guilty of similar conduct” among § 3553(a) factors). Accordingly, applying our reasoning from Myers, we conclude that the district court’s assessment of McDonald’s likelihood of recidivism based solely on his age and the USSC Recidivism Study cannot justify a wide divergence from the guidelines sentence in this case because the guidelines sentencing range already accounts for the likelihood of recidivism.
Moreover, we are unpersuaded by the district court’s use of the USSC Recidivism Study. As a preliminary matter, it appéars that the district court simply adopted Nellum’s recitation and application of the study.4 But Nellum was a much different offender. 2005 WL 300073, at *3. Nellum was 57 years old when he *955was sentenced and would not be released until he was 65. Id. at *3. McDonald was only 38 at sentencing. After serving a 132-month sentence, McDonald would be 49. In other words, at the time that McDonald would be released, McDonald would still be eight years younger than Nellum was when Nellum was sentenced.
Nor were the two defendants’ criminal histories comparable. Two misdemeanor crack possession convictions placed Nellum in his criminal history category of III. Id. at *4. McDonald, on the other hand, had a significant and steady history of committing crime. In addition to his 2004 convictions, between 1990 and 2000, McDonald was convicted of methamphetamine possession three times (in 1990, 1995 and 1996); possession with intent to deliver methamphetamine twice (in 1994 and 1995); marijuana possession (in 1995); various theft offenses three times (once in 1994 and twice in 1995); providing false information to the police (in 1995); driving while intoxicated (in 1993); driving while barred (in 2000); and driving on a suspended license three times (twice in 1994 and once in 1996). Starting with his first felony conviction at age 24 and including the instant offenses, McDonald was convicted of 18 different crimes over the succeeding 13-year period. McDonald’s convictions resulted in a variety of sentences that included jail time, probation and court-ordered substance abuse treatment. In several instances, McDonald’s probation was revoked (or he was jailed for contempt of court) because he offended yet again.
Moreover, unlike Nellum, McDonald was a career offender. The Sentencing Commission created the career offender provision, Guidelines § 4B1.1, in response to Congress’s directive to “assure that the guidelines specify a sentence to a term of imprisonment at or near the maximum term authorized for categories of defendants in which the defendant is eighteen years old or older and ... has previously been convicted of two or more prior felonies, each of which is ... an offense described in [21 U.S.C. § 841].” 28 U.S.C. § 994(h); see also United States v. Bell, 445 F.3d 1086, 1088-89 (8th Cir.2006) (discussing the relationship between 28 U.S.C. § 994(h) and Guidelines § 4B1.1). Even without career offender status, McDonald would have faced a guidelines sentencing range of 168 to 210 months. See n. 2, supra. . Thus, the district court’s extraordinary variance down to 132 months not only negates Congress’s intent that career offenders serve sentences “at or near the [statutory] maximum term authorized,” 28 U.S.C. § 994(h), it also represents a significant downward variance from McDonald’s guidelines range without the career offender enhancement.
Even if the USSC Recidivism Study could be deemed to shed light on the likelihood that McDonald will recidivate (beyond his criminal history computation), the district court misinterpreted and misapplied it. Contrary to the district court’s apparent interpretation of the USSC Recidivism Study, the study in fact supports the guidelines’ determination that offenders like McDonald with a criminal history category of VI recidivate more often than not. According to the USSC Recidivism Study, 51.3 percent of individuals with McDonald’s criminal history category, VI, and age at sentencing, 38, recidivated within two years of release. USSC Recidivism Study at 28 (Exhibit 9). Offenders under the age of 21 who were in McDonald’s criminal history category of VI recidivated 55 percent of the time — only 3.7 percent more often than offenders who were McDonald’s age. Id. The relevant statistics stand in distinct contrast to the district court’s mistaken estimation that someone of McDonald’s age really only had a 29.4 percent chance of recidivating' — a conclusion reached by mistakenly using the sta*956tistics relating to a defendant with a criminal history category of III like Nellum, rather than VI like McDonald.5
Even within those categories, the USSC Recidivism Study noted that characteristics other than age influence the likelihood of recidivism. Those who lacked “stable employment in the year prior to their instant offense” and those who used “illicit drugs within one year prior to their instant offense have a higher recidivism rate.” Id. at 12-13. McDonald fits both of these categories. At bottom, the USSC Recidivism Study does not support a variance in this case.
We do not rule out the possibility that some extraordinary work history or individual characteristic related to age or the likelihood of recidivism could justify a variance in other cases. However, in this case, McDonald’s work history and the district court’s estimation of his likelihood of recidivism — whether considered separately or together — do not constitute the type of compelling justifications necessary to justify a variance of the magnitude awarded here.6 Accordingly, we conclude that the district court “commit[ted] a clear error of judgment by arriving at a sentence that lies outside the limited range of choice dictated by the facts of the case.” Haack, 403 F.3d at 1004.7
*957III. CONCLUSION
For the foregoing reasons, we vacate McDonald’s sentence as unreasonable and remand for resentencing consistent with this opinion.
. "A defendant is a career offender if (1) the defendant was at least eighteen years old at the time the defendant committed the instant offense of conviction; (2) the instant offense of conviction is a felony that is ... a controlled substance offense; and (3) the defendant has at least two prior felony convictions of ... a controlled substance offense.” Guidelines § 4B1.1.
. Pursuant to Guidelines § 4B1.1, every career offender's criminal history category is VI. As the district court noted, even without career offender status, McDonald’s criminal history category would have been V. Had McDonald not received a career offender enhancement, his total offense level (after an adjustment for acceptance of responsibility) would have been 31, and his guidelines sen*951tencing range would have been 168 to 210 months.
. Nellum refers to United States Sentencing Commission, Measuring Recidivism: The Criminal History Computation of the Federal Sentencing Guidelines (2004), available at http://www.ussc.gov/publicat/Recidivism_Gen-eral.pdf [hereinafter the "USSC Recidivism Study”].
. The USSC Recidivism Study contains an exhibit detailing the percentage of offenders that recidivate within two years of release. USSC Recidivism Study at 28 (Exhibit 9). The exhibit categorizes offenders using two variables: age at sentencing and criminal history category. It contains one column for each of the six criminal history categories and rows that reflect offender age categories. Nellum, upon which the district court relied, reprinted only one column of the six-column chart — a column that set forth the recidivism rates for offenders with a criminal history category of III. The district court clearly relied upon the limited portion of the . study reproduced in Nellum. Sentencing Tr. at 27-28 (“Judge Simon [from the Northern District of Indiana] quotes the United States Sentencing Commission's own report .... As the report states, quote .... According to the history category III Defendants like Defendant Nellum the recidivism rates are as follows .... ”).
. According to Exhibit 9 of the USSC Recidivism Study, 19.8 percent of offenders with a criminal history category of III who were over age 50 at sentencing (like Nellum) recidi-vated within two years. USSC Recidivism Study at 28. Of those offenders with a criminal history category of III who were between 36 and 40 years of age at sentencing, 29.4 percent recidivated within two years. Id. However, 51.3 percent of offenders with a criminal history category of VI who were between 36 and 40 years of age at sentencing (like McDonald) recidivated within two years. Id.
. We make no holding concerning whether the record in this case could support a downward variance of some lesser magnitude. The district court was unwilling to vary based upon the death of McDonald’s daughter, as it was unable to determine if, as McDonald urged, the death of McDonald’s daughter really was the triggering event for McDonald's relapse into drug use. As the district court stated, it was "in no position to know" whether there was a causal connection between the death of McDonald’s daughter and McDonald’s relapse.
The district court was less clear, however, with respect to McDonald’s other personal characteristics. Specifically, the district court’s ambiguous statement that "[s]pecific mention by me will be [made] of the Defendant's long-time drug dependency and psychiatric diagnoses” is equally consistent with the district court's consideration and dismissal of those characteristics as it is with consideration and variance based upon them. Given the district court's detailed reliance on McDonald’s work history and the recidivism study, we cannot merely assume that the district court based its variance upon McDonald’s addiction or mental health. Cf. United States v. Gatewood, 438 F.3d 894, 896 (8th Cir.2006) ("For this court to properly carry out the appellate review mandated by Booker, it is essential that the district court explain the reasons why it has imposed a sentence outside the guidelines sentencing range in a particular case.”). In any event, the district court’s ambiguous statements regarding McDonald's addiction and mental health do not support a variance of the magnitude awarded.
.The dissent, following United States v. Meyer, 452 F.3d 998, 1000 n. 3 (8th Cir.2006) (Heaney, J., concurring), summarizes the results of our post-Booker sentencing review cases, including both departures and variances, and expresses concern that we have tended to affirm sentences imposed above the advisory guidelines range (sixteen affirmed, one vacated) and vacate those below the range (five affirmed, twenty-five vacated). We accept the dissent's invitation "to take a critical look at how we dispose of such cases.” Post at 960. A closer look at the cases
listed in Meyer, as supplemented by the dissent, reveals several mundane explanations for the disparity.
First, the dissent's list includes seven sentences varying downward that were vacated by necessity because the sentencing court did *957not explain any reasoning for the downward variance. Shafer, 438 F.3d at 1227 (“For this court to properly carry out the appellate review mandated by Booker, it is essential that the district court explain why it imposed a sentence below the guidelines range .... ”). This should occur with less frequency as sentencing courts have begun to comply with our only recently developed Booker-review jurisprudence. Indeed, sentences vacated for absence of reasoning might well be affirmed when they appear before us again with the benefit of the sentencing court's reasoning.
Second, the dissent’s list includes seven sentences (five of which are not included in the previous paragraph) that depart or vary downward more than 100 months, but only one that departs or varies upward by that amount. Because extraordinary departures and variances must be supported by extraordinary circumstances, McMannus, 436 F.3d at 874, it stands to reason that departures and variances of the largest magnitude have less of a likelihood of surviving appellate review.
Third, the list includes five sentences that were vacated where the sentencing court imposed no prison term at all, despite a lower end of the advisory guidelines range of 12, 24, 30, 46 and 63 months respectively. (In addition, one sentence varying downward from a low end of 12 months to no prison term was affirmed.) We have suggested that a variance to zero prison time where the Sentencing Commission has found that substantial prison time is indicated also requires extraordinary justification. United States v. Gall, 446 F.3d 884, 889 (8th Cir.2006).
Finally, and most tellingly, statistics published by the Sentencing Commission establish that, between the publication of Booker and February 2, 2006, sentencing courts in the Eighth Circuit pronounced 767 below-guidelines range sentences (169 departures and 598 variances; not counting those sponsored by the Government) and 94 above-guidelines, range sentences (17 departures and 77 variances), a ratio of more than 8:1 below-guidelines to above-guidelines sentences. See United States Sentencing Commission, Final Report on the Impact of United States v. Booker on Federal Sentencing D-20 (March 2006). Although the time frame in the dissent's list of cases is not an exact match, the ratio of below-guidelines range sentences appealed to above-guidelines sentences appealed in the list is less than 2:1. These statistics indicate that about three-quarters of below-guidelines range sentences do not reach us for appellate ■ review and are, thus, effectively affirmed, although not eligible for the dissent’s list. Furthermore, one would expect that the limited number of below-guidelines sentences actually appealed by the Government represent those that appear to be'the most debatable, and, thus, the most likely to be overturned on appeal.
Based on this non-exhaustive list of factors, it appears that the discretion afforded to sentencing judges post-Booker is not "an escalator that only goes up.” Post at 960 (quoting United States v. Jones, 460 F.3d 191 (2d Cir. 2006)).